Research on Improved Transformer Fault Diagnosis Method Driven by IBKA-VMD and Hierarchical Fractional Order Attention Entropy Synergy
Open Access
- 16 March 2026
- journal article
- research article
- Published by MDPI AG in Fractal and Fractional
- Vol. 10 (3) , 195
Abstract
Rolling bearing faults are the primary cause of rotating machinery failure. Under complex operating conditions, the weak fault impact signals are easily overwhelmed by strong noise and exhibit significant non-stationary characteristics, posing severe challenges to accurate diagnosis. To address this, this paper proposes an improved Transformer-based fault diagnosis method driven by the improved black-winged kite algorithm-variational mode decomposition (IBKA-VMD) and hierarchical fractional-order attention entropy (HFrAttE). The method employs the integrated multi-strategy IBKA to adaptively determine the optimal parameters of VMD, utilizes HFrAttE to construct highly discriminative feature sets, and further builds an improved Transformer model integrating bidirectional attention mechanisms and feature decoupling structures for deep feature mining. The classification decision is finalized by the twin extreme learning machine (TELM). Experimental results on the case western reserve university (CWRU) bearing dataset under different noise environments (−2 dB, −5 dB) demonstrate that the proposed method maintains 100% accuracy, recall, and F1-score under −5 dB noise interference, significantly outperforming comparative models. It exhibits excellent anti-noise performance and feature extraction capability, providing an efficient solution for intelligent operation and maintenance of rotating machinery under complex operating conditions.Keywords
Funding Information
- Yunnan Fundamental Research Projects (202301AT070256)
- Training Program for Baoshan Xingbao Talents (202303)
- 10th batches of Baoshan young and middle-aged leaders training project in academic and technical (202109)
- Zhejiang Provincial Natural Science Foundation of China (LQN26F030049)
- Project of Quzhou Science and Technology Plan (2025K232)
This publication has 30 references indexed in Scilit:
- Range-free localization algorithm based on modified distance and improved black-winged kite algorithmComputer Networks, 2025
- A study on rolling bearing fault diagnosis using RIME-VMDScientific Reports, 2025
- An innovative complex-valued encoding black-winged kite algorithm for global optimizationScientific Reports, 2025
- A novel multi-sensor local and global feature fusion architecture based on multi-sensor sparse Transformer for intelligent fault diagnosisMechanical Systems and Signal Processing, 2024
- Transformer fault diagnosis method based on the three-stage lightweight residual neural networkElectric Power Systems Research, 2024
- Motor fault classification using hybrid short-time Fourier transform and wavelet transform with vibration signal and convolutional neural networkJournal of the Brazilian Society of Mechanical Sciences and Engineering, 2024
- Black-winged kite algorithm: a nature-inspired meta-heuristic for solving benchmark functions and engineering problemsArtificial Intelligence Review, 2024
- Diagnosing Faults in Rolling Bearings of an Air Compressor Set Up Using Local Mean Decomposition and Support Vector Machine AlgorithmJournal of Vibration Engineering & Technologies, 2024
- Vibration-based damage identification in composite plates using 3D-DIC and wavelet analysisMechanical Systems and Signal Processing, 2022
- Twin extreme learning machines for pattern classificationNeurocomputing, 2017